Design Paerns for Data-Driven News Articles
Shan Hao
Shanghai University
Shanghai, China
hshan112592@gmail.com
Benjamin Bach
University of Edinburgh
Edinburgh, UK
bbach@ed.ac.uk
Zezhong Wang
Simon Fraser University
Vancouver, Canada
wangzezhong2016@gmail.com
Larissa Pschetz
University of Edinburgh
Edinburgh, UK
L.Pschetz@ed.ac.uk
Context enhancing
Headline
Stating an issue
Headline
Making an evaluation or Judgment
Headline
Asking a question
Headline
Sensationalism
Data Source
Giving the link directly
News Lead
Image lead
News Lead
Textal lead
Headline
Data highlighting
Data Source
Mentioning the data source
Begin
Middle End
Narrative - Structure
Freytag’s pyramid
Narrative - Structure
Drilling-down
Narrative - Structure
Compositing
Narrative - Data-driven
Stating the value
Narrative - Data-driven
Comparing between regions
Narrative - Data-driven
Contrasting
Narrative - Data-driven
Changing over time
Narrative - Structure
Inverted pyramid
Narrative - Data-driven
Data notables
Narrative - Context enhancing
Citing the authority
Narrative - Context enhancing
Co-relation and causality
Narrative - Context enhancing
Domain knowledge
Narrative - Context enhancing
Adding detailed data
Narrative - Context enhancing
Citing different viewpoints
Narrative - Context enhancing
Interpretation
Y
N
Narrative - Context enhancing
Inferences
Narrative - Context enhancing
Basic description
Narrative - Context enhancing
Adding emotional factors
Visualization Technique
Map
Visualization Technique
Stacked chart
Visualization Technique
Tree maps
Visualization Technique
Slope chart
Visualization Technique
Bubble chart
Visualization Technique
Scatter plot
Visualization Technique
Pictogram
Visualization Technique
Combined charts
Visualization Technique
Other
Visualization Caption
Visualization attribution
Visualization Caption
Scope of data collection
Visualization Caption
Describing the content of the
visualization
Visualization Caption
Additional information
Visualization Annotation
Textual annotations
Visualization Annotation
Threshold annotations
10
Visualization Annotation
Numerical annotations
Visualization Caption
Mentioning the data source
Visualization Annotation
Visual annotations
Visualization Title
Data highlighting
Visualization Title
Stating an issue
Visualization Title
Making an evaluation or judgment
Visualization Title
Asking a question
Visualization Technique
Table
Visualization Technique
Bar chart
Visualization Technique
Line chart
Image
Illustration
Visualization Technique
Small multiples
Visualization Interaction
Selecting
Visualization Interaction
Zooming
Visualization Interaction
Exploring
Visualization Interaction
Scrollytelling
Video
Data video
Visualization Interaction
Other
Visualization Interaction
Connecting
Visualization Interaction
Inspecting
Video
Complementary Video
Narrative - Context enhancing
Presenting a different statistic
method
Narrative - Context enhancing
Data analysis
Narrative - Context enhancing
Predicting the future trend
Narrative - Context enhancing
Describing the content of
the visualization
Narrative - Context enhancing
Summarize
Narrative - Context enhancing
Linking to external references
Narrative - Context enhancing
Actions
Data
Narrative - Context enhancing
Data source description
Image
Photo
Figure 1: Design patterns (left) and the corresponding use identied from a data-driven news article example (right).
ABSTRACT
Technological advancements have resulted in great shifts in the pro-
duction and consumption of news articles. This, in turn, lead to the
requirement of new educational and practical frameworks. In this
paper, we present a classication of data-driven news articles and
related design patterns dened to describe their visual and textual
components. Through the analysis of 162 data-driven news articles
collected from news media, we identied ve types of articles based
on the level of data involvement and narrative complexity: Quick
Update, Brieng, Chart Description, Investigation, and In-depth
Investigation. We then identied 72 design patterns to understand
Permission to make digital or hard copies of all or part of this work for personal or
classroom use is granted without fee provided that copies are not made or distributed
for prot or commercial advantage and that copies bear this notice and the full citation
on the rst page. Copyrights for components of this work owned by others than the
author(s) must be honored. Abstracting with credit is permitted. To copy otherwise, or
republish, to post on servers or to redistribute to lists, requires prior specic permission
and/or a fee. Request permissions from permissions@acm.org.
CHI ’24, May 11–16, 2024, Honolulu, HI, USA
© 2024 Copyright held by the owner/author(s). Publication rights licensed to ACM.
ACM ISBN 979-8-4007-0330-0/24/05
https://doi.org/10.1145/3613904.3641916
and construct data-driven news articles. To evaluate this approach,
we conducted workshops with 23 students from journalism, de-
sign, and sociology who were newly introduced to the subject. Our
ndings suggest that our approach can be used as an out-of-box
framework for the formulation of plans and consideration of details
in the workow of data-driven news creation.
CCS CONCEPTS
Human-centered computing
Information visualization;
General and reference Design.
KEYWORDS
Design Patterns, Data Journalism, Data-Driven Storytelling, Educa-
tion, Classication
ACM Reference Format:
Shan Hao, Zezhong Wang, Benjamin Bach, and Larissa Pschetz. 2024. De-
sign Patterns for Data-Driven News Articles. In Proceedings of the CHI
Conference on Human Factors in Computing Systems (CHI ’24), May 11–
16, 2024, Honolulu, HI, USA. ACM, New York, NY, USA, 16 pages. https:
//doi.org/10.1145/3613904.3641916
CHI ’24, May 11–16, 2024, Honolulu, HI, USA Hao et al.
1 INTRODUCTION
Data journalism integrates traditional journalism with analytics ap-
proaches like data gathering, cleaning, analysis, visualization, and
publication [
31
]. Data-driven news articles are artefacts that blend
the textual narrative with data and visualizations and have recently
become a pillar in the contemporary news landscape. Currently,
the creation of such articles is now shaped by the emergence of
digital technologies and new methods of dissemination, accommo-
dating both detailed and selective reading habits, for instance, by
emphasizing key excerpts [
1
]. Furthermore, the landscape of article
production and consumption is undergoing profound shifts due
to the swift expansion of data resources, a diverse array of data
mining technologies, the establishment of communal databases,
and worldwide collaborations [
26
,
49
]. Such advancements have
given rise to what can be termed as ‘big data articles’, marking
a signicant evolution in the eld [
10
,
38
]. Eventually, there are
many ways to design data-driven articles, e.g., by integrating dash-
boards [
4
] and interactive visualizations, leveraging storytelling
and other methods from data-driven storytelling [19, 59].
Because of this, designing rich and informative data-driven arti-
cles presents a complex challenge; Creators are required to navigate
a multitude of design decisions, encompassing genre conventions,
data visualization, interactivity, depth of analysis and explanation
of visualization and data, including strategies such as highlighting
the main messages in the visualizations. While existing research,
notably on visual elements and interaction techniques [
74
], presen-
tation frameworks [
43
,
53
], and key features [
69
] of data-driven
new articles tend to focus on long-form stories or award-winning
projects. Shorter articles and the more daily routine practices have
not received adequate exploration. Hence, many educators and
novices still do grapple with communicative skills for eectively
integrating data visualization and news articles. A lack of under-
standing of these components and their combination makes it hard
to make deliberate design decisions.
In this paper, we provide a taxonomy of article types alongside a
set of design patterns to describe the richness of data-driven news
articles and inform creators’ design choices. By analyzing articles
from 6 major news media websites, we were able to classify data-
driven news articles into ve categories: Quick updates, Briengs,
Chart Descriptions, Investigations, and In-Depth Investigations. This
framework is intended to facilitate determining the communication
goal, the scope of information, and the emphasis when composing
an article. Next, we identied 72 design patterns across 11 groups,
each pattern describing common components and solutions for data-
driven news articles. Besides their analytical value, these design
patterns can serve as an educational framework for structuring and
presenting information in articles in a modular and scalable design
approach.
Eventually, we conducted interviews with 7 experts and edu-
cators in the eld of data-driven storytelling and journalism, and
evaluated the design patterns in workshops with 23 participants
to create outlines for data-driven news articles. The outcomes
demonstrated that identifying the article types and design patterns
can assist participants in understanding the nuance of structure
and opens up a broader spectrum of creative opportunities for
data-driven news article construction. A detailed description of all
our design patterns, and the workshop can be accessed online at:
https://datadrivenarticle.github.io/
2 RELATED WORK
A data-driven news article is a data journalistic artefact that presents
a journalistic data story [
69
] in a magazine style (in the taxonmy by
Segel and Heer [
59
]). In the following, we review literature about
classication, creation process, and creation support for data-driven
news articles.
2.1 Classication of Data Journalism
With the dawn of the digital age, data journalism has witnessed
a transformative journey. It began with the advent of computer-
assisted reporting in the 1950s [
23
,
37
], gained momentum with
the implementation of the Freedom of Information Act and re-
lated legislation from 1960s [
58
], conceptualized in 2008 [
27
], and
has now ourished into a commonplace practice of data journal-
ism [
10
]. Global news organizations have undertaken extensive
data journalism projects with variations in their structure and use
of visualization [
68
]. Faced with this diverse landscape of practices,
describing types of data-driven articles can benet teaching and
comprehension of this form of news [
57
], help meet the diverse
reading preferences [
1
], assist creators in selecting projects aligned
with their interests and capabilities [
68
], and can oer templates
and structured references for recurring practical projects [48].
Several taxonomies have been developed to describe the charac-
teristics and production practices in data journalism. For example,
Uskali et al. [
67
] sub-classify data-driven journalism into investiga-
tive data journalism, general data journalism, and real-time data
journalism, based on criteria such as production time, story scope,
technical level, data set, data validation, and analysis method. Simi-
larly, Borges-Rey [
9
] describes data journalism practice in the UK
as either a) daily, with a quick turnaround, generally-visualized,
and brief, b) extensive, thoroughly researched and investigative; or
c) light, editorialized, entertaining, often-humorous, and gamied.
Google News Lab’s classication [
57
] focuses on the relation be-
tween story and data and describes three types—Stories Enriched by
Data in which data is used to verify the report; Stories Using Data
to Investigate which surfaces stories hidden in the data; and Stories
Explaining Data that focus on providing meaning behind the data.
All these classications help establish a high-level understanding of
data journalism and its production. Yet, they pay limited attention
to the form of data-driven news articles, their content, their the
construction, and the use of data visualization. Our classication
of article types and their design patterns lls this gap by providing
specic solutions, ideas, and templates.
2.2 Data-Driven Story Creation Process
Several models describe the creation process of data-driven stories.
For example, Kosara et al. [
39
] develop a basic model based on
journalistic practices from information collection to presentation.
Lee et al. [
42
] describe three main components of data-driven story-
telling process—exploring data, making a story, and telling a story.
Building on this work, Chevalier et al. [
16
] describe artifacts and
roles involved in the storytelling process and note that crafting a
story and constructing story material require dierent skill sets.
Design Paerns for Data-Driven News Articles CHI ’24, May 11–16, 2024, Honolulu, HI, USA
Similarly, the production process in data journalism is described as
getting data, understanding data, and delivering data [
10
], and by
Mirko Lorenz [
12
] as four stages—compile (i.e., start with either a
question that needs data, or a dataset that needs questioning), clean,
context, and combine. This creation process is rarely always linear
and journalists adopt a multi-faceted, iterative, and often cyclical
process to news creation [
61
]. Our work primarily concentrates on
the stage of narrative construction, providing materials for teach-
ing, and designing data-driven news articles, rather than oering
technical support for specialized tasks within those stages, such as
data analysis or creating data visualization.
2.3 Data-Driven News Article Creation Support
Educational materials and creation support for data journalism in-
clude handbooks with the introduction of journalistic practice in
newsrooms with case studies [
10
,
27
], models of narrative structures
of news articles (e.g., [
18
,
33
]), and the key elements in data journal-
ism [
18
,
69
,
74
]. Regarding specic approaches for constructing jour-
nalistic stories, McKane [
46
] identies the in medias res approach,
which starts with the most dramatic moment and then moves on
to the background and less important details (also known as the
inverted pyramid structure [
50
]). Another option is the chronologi-
cal approach which is rooted in Aristotle’s Tragedy Structure: the
beginning, middle, and end [
22
]. This basic structure is further ex-
tended by Freytag [
24
] as Exposition, Rising action, Climax, Falling
action, and Resolution, has found application in data stories and
journalism [
20
,
71
]. Other models provide various strategies for
organizing the content in investigative news articles, such as the
Martini Glass [
59
], the Kabob [
32
], the Stack of Blocks [
25
], and the
Water Tower [33].
In addition to structure, design guidance introduced in general
data stories and visualizations can serve as valuable inspiration
for crafting data-driven news articles. Examples include genres of
narrative visualization [
47
,
59
], structures for sequencing [
35
], the
interplay between text and visualizations [
41
], or story synthesis
from visual analytics [
15
]. Some studies have further supported the
education and creation of data storytelling by describing design
space and patterns, such as narrative patterns for data-driven story-
telling [
6
], design patterns for data comics [
7
], narrative semantics
of Data Videos [
60
], dashboard [
4
], embellishment [
14
] and compo-
sition [
36
] for data visualizations. The creation supports mentioned
above provide invaluable insights, including a high-level workow,
guidance for individual components and specic techniques, and
specic tools or automation techniques. However, as these high-
level theoretical categories fall short in oering practical guidance,
a gap remains for novices in the creation of data-driven news arti-
cles. Our approach provides a practical approach, intended to assist
creators in informing the structure of data-driven news articles and
to bridge the gap between theoretical knowledge and practical use,
particularly for novices to data-driven articles.
3 METHODOLOGY
Our research methods included 1) article sampling, iterative coding
of 2) article types and 3) design patterns, as well as user evaluation
through 4) one-on-one sessions and 5) workshops (Fig. 2).
3.1 Article Sampling and Article Coding
Our initial sample of articles focused on the global outbreak of
COVID-19 (Stage 1, Fig. 2). As an international emergency, COVID-
19-related news reports provided a conned topic, lots of data
reporting, and allowed comparison among those articles. We con-
ducted our sampling from 6 prominent English-language news me-
dia, chosen for their high website trac, dedicated data columns,
diverse editorial styles, and global reputation for representative
journalism. All of these media outlets cater to a wide range of reader
demographics [
63
] and political inclinations [
30
], providing a rich
and varied sample of data-driven news articles. The data-driven
news articles were sampled from The Guardian (n = 26), The Econo-
mist (n = 20), The Times (n = 17), Financial Times (n = 18), New York
Times (n = 12), and BBC News (n = 9), and were randomly sampled
from the respective COVID-19 related columns. In the process of
sampling, we ensured that the collected data-driven news articles
were presented in magazine style [
59
] , and contained at least one
data visualization each. All 102 samples were published between
March 2020 and April 2022.
Through coding these articles into dierent types, we aimed to
understand how the evolving production and consumption land-
scape inuences the presentation of these articles. This aids in
comprehending their structural layout and positioning within the
broader landscape. Our approach was inspired by the work of
Uskali [
67
] and Rogers [
57
], who categorized various data parame-
ters including data sources, quantity, journalistic verication, pro-
duction cycle, and usage. However, our focus was primarily on the
types of articles rather than the process of their creation. As such,
we adapted their classication method by omitting elements like
production cycle and technical level.
We classied articles by variation of ve criteria (detailed in
Section 4) and employed both inductive and open coding methods
(Stage 2, Fig. 2). Two coders coded the same 15 articles indepen-
dently and improved the codes based on three rounds of discussions
until an agreement was reached. Then, one of the authors completed
the coding of all 102 articles. The coding resulted in labels describ-
ing the level of ve criteria by describing how each criterion is
presented in an article, which was then used to classify data-driven
news articles into 5 types (Section 4).
3.2 Coding for Design Patterns
In this second coding stage (Stage 2) we coded articles according to
common design pattern for an article’s syntactic components and
their visualizations: headline, data sources, narratives, visualization
titles, visualization techniques, visualization annotations, visualiza-
tion captions and interactions. To prevent uniformity, we extended
the set of articles for this coding stage to topics other than COVID-
19. First, we randomly chose 4 articles about COVID-19 articles
from each article type from our initial set of 102 articles. Second,
we went back to the 6 media outlets and sampled new data-driven
news articles from the respective columns on economy, society, cul-
ture, politics, health, and others (Fig. 3). However, through random
sampling, we found that many articles belong to our type investi-
gations and hence do not suciently cover our other four article
types. In the hope of increasing article type coverage, we searched
for other articles by individual authors named on these randomly
CHI ’24, May 11–16, 2024, Honolulu, HI, USA Hao et al.
Stage 1 Stage 2 Stage 3
Financial Times
The Guardian
The Times
BBC news
The Economist
New York Times
News Media
Articles From Other Topics
Coding Coding
11 Common
Components
60 Design Patterns
Stage 4
Stage 5
With Experts
One-on-one
sessions
With Educators
Interview
(Offline+Online)
Workshop
Iterating
8 Students
Offline Workshop
15 Students Educators
Online Workshop
5 Types of Articles
72 Design Patterns
5 Types
Common Components
Design Patterns
*4
*4
*7
*3
Other Topics
60
COVID-19 Topic
20
COVID-19 Topic
102
Sample From COVID-19 Topic
Add
Data-Driven News
Article
Sampling
of Articles
Figure 2: Overview of study methods. Data-driven news articles were sampled and coded twice, once for article types (Stage 2)
and again for design patterns (Stage 3). Stage 4 involves soliciting feedback for improvements from experts. In Stage 5, we
conducted workshops and interviewed educators to evaluate the use of our design patterns.
Financial Times(FT)
Covid-19 Topic Other Topics
0 5 10 15 20
New York Times(NYT)
The Economist
The Guardian
The Times
BBC
covid-19
0 5 10 15 20
economy
politics
society
environment
sport
royal life
culture
population
education
language
health
A
B
Figure 3: Frequency of (A) sources and (B) topics in the 80 data-driven articles sampled in Stage 3 (Fig.2)
sampled articles. Our assumption was that these individual articles
might be shorter while still published by the same media outlet.
These individual authors identied their roles as ‘visual project
editor, ‘data project, visual storytelling reporter, and ‘information
designer. In total, this gave us 35 articles. Together with 25 of the
randomly sampled team-authored articles, and the 20 articles from
the COVID-19 topic, our sample set for Stage 3 coding was 80. Last,
we coded all of those new 60 articles according to our article types
(Stage 3) to verify we did not encounter any new article types. All
those articles were published between 2019 to 2022 and followed
the same selection criteria as our initial set of 102 articles. Across
all 80 articles, our samples contained 1,224 paragraphs and 313 visu-
alizations, corresponding to 15.3 paragraphs and 3.9 visualizations
per article.
The same person who coded all 102 COVID-19 corpus coded
design patterns for each of these components and engaged in discus-
sions with the other co-authors to rene the codes. Once saturation
in renement was reached, the same person coded all 80 articles. In
total, we identied 60 design patterns (Figure 1, details in Section
5).
3.3 Iteration and Evaluation
To obtain initial feedback on the classication of the article types
and design patterns (Stage 4, Fig. 2), we conducted one-on-one ses-
sions with one data journalist (14 years experience as a professional
journalist) and three data-driven storytelling and visualization re-
searchers (2 PhD students and 1 post-doctoral researcher). In the
one-on-one sessions, we described the type of article to the expert
Design Paerns for Data-Driven News Articles CHI ’24, May 11–16, 2024, Honolulu, HI, USA
and showed relevant article cases, as well as our design pattern
cards, then invited the expert to use the cards to construct a sim-
ple article framework, which helped the expert discover problems
and provide feedback. Afterward, we asked the expert for their
experience in design articles, how our approach could be used in
their workow, and potential improvements. According to the feed-
back, we modied some names and descriptions of our article types
and patterns to enhance clarity. For instance, the article type that
provides summary information was renamed from Daily News to
Brieng to avoid giving audiences the impression that it was solely
based on information from the same day. We added narrative struc-
tures from existing literature [
18
,
33
] to serve as inspiration for
narrative construction. We have also added a few design patterns
based on their suggestions. For example, Inference was split from
the original coding as part of the interpretation. After rening the
codes we recorded all articles again. Lastly, we evaluated the design
patterns in multiple workshops with students from journalism, and
design backgrounds and interviewed their lectures about potential
future usage of involving our design patterns in their classrooms
(Stage 5, Fig. 2).
4 FIVE TYPES OF DATA-DRIVEN NEWS
ARTICLES
We identied ve types of data-driven news articles in Stage 2 of
our process. We categorized articles based on ve key criteria:
C1:
DataThe range of data sources from single to multiple,
collection techniques from public and open access data sets
or unocial information sources and self-collected data, and
analytical methods.
C2:
Analysis—The depth of data description, ranging from
overview summaries to detailed explanations, and whether
visualizations are employed to support the data.
C3:
ContextThe degree of contextual detail provided, from
minimal to extensive.
C4:
InvestigationThe breadth and depth of the investi-
gation, topic, and factors associated with the event, from
narrow to wide-ranging.
C5:
DiscussionThe depth of discussion, from supercial to
comprehensive.
We specically focus on content and structure, as these components
indicate the depth of data analysis, the logic used, and the intended
objectives of data-driven news articles. Our focus does not extend
to the authoring process in terms of the collaboration model among
journalists, the number of collaborators, technical prociency, or
production time. We also abstain from categorizing articles based
on text length or the number of visualizations. For example, Maga-
zine Dashboards [
4
], could be longer if they display multi-faceted
information with many visualizations or shorter if they focus on a
single aspect; however, only the length and numbers of data visual-
izations do not necessarily dictate a change in the article type. Also,
the ve types are not intended to have rigid boundaries; rather, they
serve to describe ve typical representations along a spectrum.
4.1 Article Typ e: Quick Updates
Quick Updates generally oer a snapshot of current data, featuring
visualizations as the key elements and limiting textual descriptions
for quick comprehension. Similar to Uskali’s Real-Time News [
67
],
data presented in the articles are automatically updated, and Mag-
azine Dashboards [
4
] (Fig. 4A, [N1]), which provides at-a-glance
insight [
73
], Quick Updates is commonly used in journalism for
presenting data related to public health or political elections, e.g.,
data update during the COVID-19 pandemic period [
75
]. According
to the examples we coded, Quick Updates mainly use open pub-
lic data (C1), and cover multiple facets of an event with limited
text descriptions (C2), the main component of this article type is
visualizations.
4.2 Article Typ e: Briengs
Briengs provides concise summaries and overviews of events. It
typically includes a summarized overview of an event and data
over a certain period, oering audiences a glimpse of trends, or the
latest developments. For example, the article from BBC News titled
Coronavirus: What is the R number and how is it calculated? (Fig. 4B,
[B4]) presents a brief explanation of the R value, its variations
across dierent regions in the UK, and three key infectious disease
metrics. The article provides a high-level overview of the subject
without delving into detailed interpretations. Similarly, the World
This Week column in The Economist, featuring concise articles like
Business [E2], oers varied economic summaries without delving
deeply into analysis.
Briengs typically source their data from open data or data from
other published articles (C1). They provide event summaries and
highlight signicant data changes but do not delve deep into the
reasons behind the data (C2). More than just presenting minimum
descriptions as in Quick Updates, the textual descriptions in Brief-
ings are more informative, and visualizations are used as supple-
ments rather than being main components. In Briengs, the depth
of data analysis remains shallow, primarily oering a collection of
summary-level insights (C3).
4.3 Article Typ e: Chart Descriptions
Chart Descriptions indicates news articles with an explicit narra-
tive structure, depicts the development of events with analysis of
more than one aspect. Regarding data sources, Chart Descriptions
typically uses publicly available data to focus on various facets
of events (C1). While they oer observations, they provide only
limited insights (C2). There is ample relevance between the context
and the description of data, the text supplements the causes and
consequences of the development of the event (C3).
This type is akin to what Rogers referred to as stories that are
enriched by data [
57
], leaning more towards traditional reporting,
where data is used to validate foundational reporting. The example
from The New York Times titled What the BA.5 Subvariant Could
Mean for the United States (Fig. 4C, [NYT7]) explores the changes
brought about by the BA.5 subvariant of COVID-19 in the United
States, oering a multifaceted explanation of the challenges associ-
ated with this variant. The article then goes on to predict the future
trajectory of cases in the United States by drawing comparisons
with the situations in other countries.
CHI ’24, May 11–16, 2024, Honolulu, HI, USA Hao et al.
Figure 4: Examples of the ve types (Section 4) of data-driven news articles, with thumbnails showing the continuation of the
articles above. (A) Quick Update example: COVID-19 vaccine tracker: the global race to vaccinate. (by the Financial Times,
[FT2]); (B) Brieng example: Coronavirus: What is the R number and how is it calculate d? (by BBC News, [B4]); (C) Chart
Description example: What the BA.5 Subvariant Could Mean for the United States. (by the New York Times, [NYT7]); (D)
Investigation example: One million coronavirus deaths: how did we get here? (by the Guardian, [
28
]); (E) In-Depth Investigation
example: The pandemic’s true death toll (by The Economist [21])
4.4 Article Typ e: Investigations
Investigations aims to uncover critical issues or provide a deeper un-
derstanding of complex phenomena. This type of article draws on a
range of data sources, including government open datasets and aca-
demic reports, and may also include data unearthed by journalists,
to provide a well-rounded view of the issue (C1). It often involves
storytelling through data-driven narratives using visualizations
with interactions to engage and inform audiences (C2). These arti-
cles exhibit close contextual connections, and each section builds
upon the previous one in a progressive manner (C3). Journalists
compile and analyze fragments of information from diverse sources
to reveal hidden patterns, trends, and insights accompanied by di-
verse types of data visualization techniques (C4), to guide readers
to reect on the event (C5). The Guardian’s article titled One million
coronavirus deaths: how did we get here? (Fig. 4D, [
28
]) chronicles
the progression of COVID-19, leading to one million deaths in a
span of nine months. Through the creation of character sketches of
individuals who succumbed to the virus in various countries and
interviews with their family members, along with an exploration of
the overall development of the disease, the story sheds light on the
resultant policy and lifestyle changes, and the grief experienced by
the loved ones of the deceased. By investigating the personal stories
behind the death toll, the report encourages readers to contemplate
more deeply the real-life impact of the relevant data.
4.5 Article Typ e: In-Depth Investigations
The In-Depth Investigations features articles that delve much deeper
than surface-level reporting, utilizing comprehensive data analy-
ses to uncover complex trends, relationships, or anomalies. These
pieces often involve extensive research, rich data sources from
multiple channels (C1), and sophisticated statistical and data visu-
alization techniques (C2), to bring new insights through lengthy
narratives (C3). The aim is to provide a thorough understanding
of the issue at hand (C4), often challenging traditional wisdom or
revealing previously hidden aspects of a subject, thereby fostering
critical discussion and thought (C5).
An example is The Economist’s article titled The pandemic’s true
death toll (Fig. 4E [
21
]). The authors expressed skepticism regarding
ocial reports of death tolls, prompting them to collect data from
121 indicators from over 200 countries and regions and create a
machine-learning model. According to the estimation, the authors
provide a central gure of 10.2 million deaths, which was twice
the ocial estimate and even as low as one-quarter in some cases.
Using new data sources and estimation models, the article oers a
dierent perspective with novel insights.
Design Paerns for Data-Driven News Articles CHI ’24, May 11–16, 2024, Honolulu, HI, USA
5 DESIGN PATTERNS FOR DATA-DRIVEN
NEWS ARTICLES
5.1 Design Pattern Descriptions
This section describes 72 design patterns for data-driven news arti-
cles (Figure 5), identied by coding the 80 articles in Stage 3, (Fig-
ure 2). Patterns are categorized according to 11 article components:
Headline, News Lead, Data Source, Narrative, Image, Visualization Ti-
tle, Visualization Technique, Visualization Annotation, Visualization
Caption, Visualization Interaction, and Video. Percentages indicate
the frequency of articles applying a given pattern. Percentages for
each component do not sum up to 100% because multiple patterns
can be used in combination. Percentages associated with design pat-
terns for visualization titles, visualization techniques, visualization
annotations, visualization captions, and visualization interactions
indicate pattern occurrence across all articles, per article, not pat-
tern frequency per visualization. In other words, a 10% value means
10% of our 80 articles feature at least one visualization with the
respective pattern.
Headline Patterns
—Prior studies have explored the pivotal role
of headlines in news articles, along with the sentiments they con-
vey [
3
,
17
]. However, research targeting headlines in data-driven
news articles remains notably limited. According to our coding,
patterns for headline ranges from directly referring data to dierent
forms of summary.
Data highlighting (3.75%) directly species
numerical values in the headline. Stating an issue (40%) not
directly displaying data but using dataset or event names, as in G4
‘Covid in the UK: Cases, deaths... . Making an evaluation or
judgment (19.75%) states a nding from the data or an identied
issue. Asking a question (36.25%) formulates questions as head-
lines, similar to an open framework [
8
]. Sensationalism (1.25%)
employs exaggerated and dramatic language to capture attention.
News Lead Patterns
The news lead, also referred to as the
‘news lede, is situated below the headline and concisely introduces
audiences to the main points and essential information of the arti-
cle [
18
]. Textual lead (71.25%) introduces the article’s content
through text, and Image lead (63.75%) uses images, photographs,
or a combination of visuals and data visualizations (56.86%) to rep-
resent the article’s content.
Data Source Patterns
The data source is a critical factor that
inuences the credibility and transparency of a news article. Data
sources were specied by Giving the link directly (73.75%) (e.g.,
URL) to the database, or Mentioning the data source (25%) in
the text. We found that only one article in our samples does not
specify the data source.
Narrative Patterns
—Regarding narrative framework, we drew
inspiration from previous research, including the analysis of data
tasks [
2
], the recognition of semantic content within articles as
explored [
45
], and the ne-grained analysis of text narrative pre-
sented by Latif et al. [
41
], who analyzed 22 data-driven geographic
news stories, distinguishing between data-driven and embedding
text narratives. Our encoding builds upon their work, presenting
an extended version and diversity introduced by various article and
visualization types. We categorize text narratives into three distinct
types: structure, data-driven text, and context-enhancing text, we
have compiled 26 relevant design patterns for these narrative types.
Previous research has explored narrative structures in news arti-
cles, although not exclusively those that are data-driven (e.g., [
18
,
33
]). Four patterns are commonly used in our samples. Inverted
pyramid structural pattern [
50
], commonly employed in traditional
journalism and still prevalent in data journalism. This pattern starts
with the most important information and then provides background
information and conclusions. Freytag’s pyramid pattern [
24
]
consists of ve phases in a narrative: Exposition, Rising action,
Climax, Falling action, and Resolution. The Drilling-down ap-
proach transitions from a broad overview to a detailed, micro-level
analysis, oering an in-depth exploration of particular facets within
a larger subject. The Compositing structure centers on a uni-
fying theme, presenting various pieces of information in either
a complementary or parallel fashion to create a comprehensive
view. Each section independently explains an aspect of the topic.
It may also take the form of a question-and-answer format, where
questions are used as titles for subsections.
There are ve types of data-driven text patterns: Stating the
value (10%) simply presents the values or the change of the value.
Change over time (63.75%) describes the trend and comparison
of values at dierent times. Contrasting (75%) is to compare the
data, such as contrasting a whole with its parts or present rankings.
The Comparing between regions approach (30%) compares data
values across dierent geographical areas. Data notables (15%)
focuses on specic values such as outliers and extrema.
Context-enhancing text patterns include:
Basic description
(62.5%) presents a summary or an overview. Citing the author-
ity (71.25%) involves referencing government or academic reports,
as well as quoting experts to enhance the article’s credibility. Co-
relation and causality (68.75%) puts data into context, reveals the
co-relation of events, and explains reasons. As in the G12 .. . these
indicators have important implications for the level of inequality:
taxation, social spending in sectors such as health, welfare and ed-
ucation, and labor rights’. To elaborate on co-relation or causality,
creators may extend to explaining concepts in specic domain
knowledge (30%). For example, G14 introduces the problem of
drinking water quality by specifying the pollutants and their ef-
fects, ’EPA stipulates that the nitrate content per liter of water is
10 mg, but it is often exceeded. This standard is designed to pre-
vent the fetus from getting enough oxygen.
Adding detailed
data (72.5%) is to provide specic examples or more in-depth data.
Inferences (60%) is to provide appropriate inferences according
to the data. Interpretation (50%) illustrates links in the devel-
opment of things, e.g., ’Despite the additional disruption caused
by Brexit, developed economies around the world are dealing with
similar problems’ (G11). Citing dierent viewpoints (18.75%)
to present dierent opinions by citing perspectives from dierent
entities. The Adding Emotional Factors category (constituting
11.25% of articles surveyed) refers to the intentional inclusion of
emotionally resonant elements in the article to elicit empathy from
readers. For instance, in coverage of the COVID-19 pandemic, an
CHI ’24, May 11–16, 2024, Honolulu, HI, USA Hao et al.
Design Paterns for Data-driven News Article
Headline
Data highlighting Stating an issue
Making an evaluation or Judgment
Asking a question
Sensationalism
?
Visualization
Title
Visualization
Caption
Visualization
Annotation
Visualization
Interaction
Video
News
Lead
Data
Source
Line chart Bar chart
Table Small multiples
Combined charts
Inverted pyramid
Freytag’s pyramid
Map
Tree maps Stacked chart
Slope chart Pictogram
Compositing
Structure
Data-driven
Context enhancing
Stating the value Change over time
Data notables
Contrasting
Scatter plot Bubble chart
Other visual representation
Comparing between regions
Basic description
Adding detailed data
Co-relation and causality
Inferences
Interpretation
Data source description
Citing different viewpoints
Adding emotional factors
Presenting a different statistic
method
Linking to external references
Summarize
Actions
Predicting the future trend
Describing the content of the
visualization
Narrative
Image
No.
conclusion
Drilling-down
Data
Data analysis
Visualization
Technique
Domain knowledge
A
B
A
A
Headline
News Lead
Narrative
Visualization Title
Annotation
Interaction
Caption
BA
Begin
Middle End
There have been 7m-13m excess
deaths worldwide during the
pandem (E14)
Textal lead
The latest updates on coronavirus
cases, deaths and hospitalisations,
using the best available national
data. (G4)
Contents are presented in the
descending order of importance
and relevance. [50]
Based on the four major narrative
categories: Establisher (E), Initial (I),
Peak (P), and Release (R)) [24]
Uber spans about 1,200 locations,
covering 10,000 towns and cities
worldwide. (G6)
Britain is ranked only 109th for the
proportion of budget it spends on
education... (G12)
In Wales, 15 per cent of the
population were aged 65 and over in
1981 — the same as London — but
that rose to 21.3 per cent in
2021.(FT8)
...the town’s water system has
exceeded the federal legal nitrate
limit by 15 times.... (G14)
In the past four months, more than
180 journalists from 44 media outlets
in 29 countries around the world
have excavated these records. (G6)
...these indicators have important
implications for the level of
inequality: taxation, social spending
in sectors such as health... (G12)
the theme of G7 is the high
temperature in Europe, and specific
cases from France and the United
Kingdom are added for auxiliary
analysis.
Despite the additional disruption
caused by Brexit, developed
economies around the world
are dealing with similar problems.
(G11)
Others, however, back the
government’s decision to unlock
now.(G10)
She is on 20mg of morphine a day to
manage the pain, yet regularly wakes
in the night.(T4)
We used statistical patterns to create
baselines... Our materials include
databases from Berkeley
collaborations and datasets created by
Karlinsky et al....(E12)
By 2050, scientists expect
new items to have the same carbon
timestamp as medieval items. (E8)
... the areas most affected by the
disaster have announced a ban on the
use of hoses... (G13)
Photo
(G13)
(Ft14)
Illustration
The UK’s lack of investment in
education and relatively low tax rates
have created a highly unequal society
in which the poor are “often unable to
cover living costs”,... (G12)
In the table below, you can check
the number of deaths in the location.
(G4)
Our COVID-19 tracking shows that Western
European countries have been slow to the
vaccine in early 2021, leading to an increase
in excess mortality. But by June the mortality
rate in the region has normalized
(E12)
the current estimate of the number of
people infected with the new crown in
the UK is based on NHS household
surveys and test samples. Times are
posted weekly.(G4)
EPA stipulates that the nitrate content
per liter of water is 10 mg, but it is often
exceeded. This standard is designed to
prevent the fetus from getting enough
oxygen. (G14)
....than most other rich countries,”said
Max Lawson, head of inequality policy
at Oxfam International.. (G12)
This discrepancy mirrors a global
trend: low-income countries tend to
have more progressive tax
structures... (G12)
Guide the reader from a wide view
to a focused view.
Each module independently
explains a portion of the theme,
Contains structures such as
Kabob, the Stack of Blocks......
Giving the link directly Mentioning the data source
Official figures say there have been
55,000 covid deaths in South Africa
since.....(E14)
Image lead
Covid in the UK cases, deaths and
vaccinations – the latest numbers
(G4)
Data highlighting Stating an issue
United Kingdom - Daily number of
new coronavirus hospitalizations
(G4)
July 2022 was one of three warmest
Julys on record (FT7)
Mentioning the data source Visualization attribution
Data: data. gov. uk (G9)
Guardian Graphic (G9)
Numerical annotations Threshold annotations
Mark numerical information in the
vicinity of a location. (B1)
Delineate intervals or values with
specific significance. (T5)
Show the specifics of the data
[hover line graph,hover line graph
dot, hover map] [11]
Mark something to keep track of it
[click line graph, click line graph dot,
click map, click list label], and [click
list checkbox] [11]
Show me more or less detail
[zoom in or out inside a map view
to adjust level of abstraction] [72]
Scrollytelling is one such emerging
approach where dynamic updates
to the article content are triggered
by scrolling [65]
Show related items [hover list] [10]
For example: Game (FT16);
website links(NYT3) ......
Show something else, [click
query-button] [11]
Inspecting Selecting
Zooming Exploring
Scrollytelling
Connecting
Other
Combining data visualizations,
animations, and audio narrations
(FT6) ......
Data video
Providing additional information to
enhance the the topic beyond the
data, with background information,
interviews, expert opinions, and
related coverage.(FT15)
Complementary Video
Textual annotations Visual annotations
Uk: More than 40 housed destroyed
in London after grass fires spread in
several areas. (G7)
refer to the provision of supplementary
information through interactive tooltips and
enhanced graphical imagery.
(E14)
Scope of data collection
Describing the content of the
visualization
correct as of 5 July 2021 - two weeks
before 19 July - allowing for vaccine
become effective. (G9)
Additional information
Note: Charts show 14-day case averages, and
the frequency of variants among cases is an
estimate. Sequencing rates can reflect
localized trends based on testing from a
particular region or hospital. (NYT7)
3-month moving average of median
hourly wage growth (FT20)
By July 2022, the burned area of the
EU is 515,000 hectares, which is four
times the estimated level recorded
since 2006. (G7)
Hospital treatment is still a waiting
game. (T4)
The global stagflation shock of
2022. (FT20)
Does Gen Z spend too much time on
social media? (E1)
Making an evaluation or judgment
Asking a question
Another massacre. (E7)
On average, how many litres of
water do you think your household
ueses every day? (E9)
Used only when precise locations or
geographical patterns in data are
more important to the reader than
anything else. [66]
Like a scatterplot, but adds
additional detail by sizing the circles
according to a third variable. [66]
The standard way to show the
relationship between two continuous
variables, each of which has its own
axis. [66]
For example: grid charts (FT1), heat
maps(E12), matrix diagram (E13), violin
plots (FT19), timeline (G5), Sankey
diagram (FT12), and illustrations (FT16).....
Excellent solution in some instances
– use only with whole numbers (do
not slice off an arm to represent a
decimal). [66]
Perfect for showing how ranks have
changed over time or vary between
categories. [66]
A simple way of showing part-to-whole
relationships but can be difficult to read
with more than a few components. [66]
Use for hierarchical part-to-whole
relationships; can be difficult to read
when there are many small
segments.. [66]
The bar/column chart has various
variants that serve different purposes,
including:Diverging Bar Chart
(Deviation); Ordered Column Chart
(Ranking) etc. [66]
Line charts are the most prevalent
type of visualization found in the
wild. The standard way to show a
changing time series. [66]
Tables provide multiple data
comparisons and overviews and can
be visually encoded through color,
fonts and mini icons. (B1)
The top and bottom of the chart are
combined to jointly display the data
content in a certain direction. (B1)
Small multiple plots are another way
to provide comparison and
spatiotemporal overview. (B1)
...using figures collated by Our
World in Data - a collaboration...
(B1)
(FT19)
Citing the authority
... even as Downing Street staff were
breaking the government’s own
lockdown rules by throwing
parties.(link to article: Boris Johnson’s
Partygate remorse lasts all of 30
seconds) (G8)
Figure 5: The full list of our design patterns for data-driven news articles with examples. The patterns are identied by coding
80 articles in Stage 3 (Fig. 2), also informed by literature on narrative structures in journalism, and design patterns such as
description of visualization techniques [66] and visualization interaction [11, 65, 72].
Design Paerns for Data-Driven News Articles CHI ’24, May 11–16, 2024, Honolulu, HI, USA
article might not only present the statistical increase in cases but
also provide real-life examples of individuals who succumbed to the
virus. Family interviews and testimonials can serve as emotional
touchpoints, adding a human element to the raw data, thereby mak-
ing the information more relatable and impactful. Data source
description (38.75%) is to provide information about the issuing
organization, sample method, and whether the data is preprocessed
or ltered for certain reasons. Presenting a dierent statistic
method (16.25%) involves the use of novel statistical approaches to
reveal insights that may not be apparent using conventional meth-
ods of data analysis. This category aims to provide fresh perspec-
tives on existing data by employing less commonly used statistical
techniques, thereby shedding new light on a topic or issue. Data
analysis (13.75%) is to present the process of analyzing the data,
interpreting the statistics, and drawing conclusions. Predicting
the future trend (26.25%) is to predict the future trend of an event
based on the current data. Describing the content of the visu-
alization (25%) involves providing an interpretation or summary
of what the data visualization is presented. Actions (20%) focuses
on detailing the actions that have been taken or are recommended
to be taken in light of the data-driven ndings. Linking to ex-
ternal references (85%) is to include links to other articles in the
text as references. Summarize (90%) means to conclude from
the pre-mentioned data and arguments as a take-home message of
a section or the entire article.
Image Patterns
—In our analysis, we identied two image de-
sign patterns in the articles:
Photo, used in 17.5% of articles
(14 out of 80), and
Illustration, found in two instances, e.g.,
illustrations are used as separators between chapters [Ft14].
Visualization Title Patterns
have been found to use the same
patterns as for article headlines.
Visualization Technique Patterns
Through statistical anal-
ysis of 313 visualizations in the 80 articles, we found that Line
charts were the most frequently employed type (26.2%), followed
by Bar charts (13.1%), Tables (10.22%), Small multiples
(10.22%), Combine d charts (9.27%), Maps (6.07%), Tree
maps (4.79%), Stacked charts (3.51%), Slope graphs (2.24%)
and Pictogram (2.24%). A smaller proportion of visualization
techniques are also found such as Scaer plots (1.6%), Bubble
charts (1.6%) and
Other chart types (8.95%), such as grid charts,
heat maps, violin plots, and innovative illustrations.
Visualization Annotation Patterns
There are many types of
annotations by their forms. Ren et al. [
56
] conducted an analysis of
variations in line charts, bar charts, and scatter plots in data jour-
nalism visualizations, proposing visual annotation methods such as
text and shapes. Building upon this, and inuenced by the presence
of visualizations like maps and information charts in our corpus,
our analysis delved further. We identied additional annotation
types beyond numerical annotations, leading us to recode and cat-
egorize our ndings into four types: Numerical annotations
(76.92%) represent the most prevalent category, encompassing nu-
merical value near its visual representation. Threshold annota-
tions (15.2%) delineate intervals or values with specic signicance.
Textual annotations (9.47%) describe data information or pro-
vide additional summary-like information. Visual annotations
(14.79%) refer to the supplementary visualization annotated in a
visualization.
Visualization Caption Patterns
Visualization caption refers
to the textual descriptions that typically appear as footnotes be-
neath the visualization or occasionally below the title (N=2). Out of
246 charts, we found patterns including Mentioning the data
source (84.5%), Visualization aribution (60.16%) that states
the creator or source of the visualization, Scope of data col-
lection (48.78%) that refers to the contextual parameters within
which data is collected, often specifying aspects like the time range,
geographical locations, or any other conditions that dene the
boundaries for data acquisition. Describing the content of the
visualization (15.85%) is the same as the pattern in the narrative.
Lastly, Additional information (4.47%) adds extra details, such
as inuencing factors to a trend or contextual information related
to the visualization.
Visualization Interaction Patterns
Yi et al. [
72
] explored in-
teractions for analysis, and subsequently, Boy et al. [
11
] supple-
mented Yi’s work, including semantic operations such as inspection,
which have been adopted in data journalism [
74
]. In our sample
articles, 124 charts (40.13%) employed interactions, revealing eight
interaction intents: Inspecting (70.16%) reveals detailed data on
hover. Selecting (39.51%) enables audiences to opt for specic
items to view, often via drop-down menus or by highlighting par-
ticular elements. Zooming (12.1%) enable audiences to display
more or less detail. Exploring (9.68%) allows audiences to search
for information according to their preferences. Scrollytelling
(4.48%) navigates audiences to the next section through scrolling.
Connecting (3.23%) links all related elements of a certain ele-
ment for comparison. And a portion of Other (6.45%) forms of
interaction, such as games or links to anchors of contents within
the article.
Video Pattern
—Within our corpus, six articles (7.5%) incorpo-
rated videos, all of which originated from the Financial Times. We
classied videos into Data video (40%), wherein information is
conveyed as animated visual narratives with data visualizations [
60
].
Complementary video (60%), providing additional information
to enhance the topic beyond the data, such as background informa-
tion, interviews, and expert opinions.
5.2 Association Between Article Types and
Design Pattern Usage
While coding patterns, we sometimes observed trends between arti-
cle types and design patterns summarized in Figure 6. For instance,
Quick Updates often use data names as headlines patterns, while
Investigations and In-Depth Investigations used more subjective or
abstract headlines (Fig. 6 A). Narrative patterns, we found, vary
with the depth of data discussion (C5)(Fig. 6 B),e.g., Quick Updates
CHI ’24, May 11–16, 2024, Honolulu, HI, USA Hao et al.
A
B
Quick Updates
Briefings
Chart Descriptions
Investigations
In-depth Investigations
One dimension (Univariate) Two Dimensions (Bivariate) Multi-dimension
Distribution of Visualization Information Dimensions (%)
0 20% 40% 60% 80% 100%
DC
Quick
Updates
Briefings Chart
Descriptions
Investigations
In-Depth
Investigations
Quick
Updates
Briefings Chart
Descriptions
Investigations
In-Depth
Investigations
Data highlighting
Asking a question
Stating an issue
Sensationalism
Making an evaluation or Judgment
Distribution of Headline Design Patterns Distribution of Narrative Design Patterns Across Different Article Types
0
2
4
6
8
10
12
14
16
Quick
Updates
Briefings Chart
Descriptions
Investigations
In-Depth
Investigations
Articles use static graphics Articles use interaction
Distribution of Interaction design pattern (%)
0
20%
40%
60%
80%
100%
E
Basic description
13
9
10
8
10
Linking to external
references
14
13
14
11
16
Data source description
5
2
6
8
10
Stating the value
Comparing between
regions
Change over time
1-4
0
5-8
9-12
13-16
Quick Updates
Briefings
Chart Descriptions
Investigations
In-Depth Investigations
4 26
1 10
7
5
1
1 13
-
13
9
- 11
Data notables
1
2
3
5
1
Citing the authority
9
15
13
14
6
Adding detailed data
10
11
15
16
6
Summarize
13
16
15
16
12
Describing the content of
the visualization
2
3
4
3
8
Actions
3
6
7
-
-
Domain knowledge
4
4
11
3
2
Citing different viewpoints
6
6
3
-
-
Adding emotional factors
1
3
5
-
-
Predicting the future trend
7
6
7
-
-
Interpretation
9
12
10
5
4
Data analysis
-
11
-
-
-
Presenting a different
statistic method
2
9
-
-
-
Contrasting
15
5
12
13
15
2
Co-relation and causality
10
13
14
16
2
Inferences
11
12
12
11
Design pattern/Per articles
Different article types ( N=16 )
0
2
4
6
8
12
10
12
7
Number of Visualizations Across Different Article Types
?
Figure 6: (A) Distribution of Headline design patterns; (B) Distribution of Narrative design patterns; (C) Number of Visualizations;
(D) Distribution of Visualization Information Dimensions; (E) Distribution of Interaction design patterns across dierent
article types.
commonly use the Stating-the-value pattern, while Investigation
articles often employ the Comparing-between-regions and the
Data-notables pattern. Context-enhancing patterns like Linking-
to-external-references are widespread, enhancing dissemination
and visibility. Chart Descriptions and Investigations frequently em-
ploy Citing-the-authority, Adding-detailed-data, and Co-
relation-and-causality, while Data-analyis and Presenting-a-
dierent-statistic-method are exclusive to In-Depth investigations.
Regarding Visualization patterns, although there is no notable
variance in the visualization titles, the quantity and complexity of
visualizations vary. Quick Updates and Investigations include more
visualizations, with Quick Updates showing higher data density,
often revealed through hovering or other interactive means. Visu-
alization Interaction patterns increase from Briengs to In-Depth
Investigations (Fig. 6 C, D, E).
6 WORKSHOP
We conducted a workshop (Stage 5, Fig. 2) to evaluate to if article
types and design patterns can help novice journalists understand
and create data-driven news articles. The workshop also sought
to identify the challenges creators faced throughout the creation
process.
6.1 Participants
The workshops were open to students from dierent backgrounds
who were interested in crafting data-driven news articles. It did not
require any specic skills. We recruited 23 participants (16 females)
aged between 19 and 35 by advertising on online social media plat-
forms and directly contacting course instructors teaching design,
digital media, and journalism from three universities in China and
the UK. 13 students majored in journalism and communication
(1 Ph.D. student, 12 undergraduate students); 8 in visual, interac-
tion design (3 Ph.D. students, 3 master students, 2 undergraduate
students); 2 graduate students in law and ecological economics
respectively. Seven participants (30%) had previously learned data
visualization in a university course and had experience in creating
between 3 and 5 data-driven articles. The remaining participants
had read articles, but had no practical experience creating them. We
conducted one oine workshop (8 people) and 8 online workshops
(15 people), and the participants were divided into 13 groups (Stage
5, Fig. 2).
6.2 Data and Materials
We prepared workshop materials around the topic of cancer, which
allowed for diverse possible perspectives for varied story types. We
provided participants with a Brieng article on the topic of ‘The 5
types of cancer with the most deaths domestically’. Alongside, we
provided data sets on cancer as well as 4-5 data visualizations on
the proportion of new cancer cases across countries and national
cancer incidences. Participants were free to seek additional data
online.
6.3 Procedure
All workshops were run by one of the authors on Miro (https://miro.com)
(Stage 5, Fig. 2), accompanied by a dierent university course ed-
ucator each. The online workshops were with students from the
journalism, law and ecological economics disciplines. The oine
workshops were with students from design discipline. After the
workshops, we interviewed three of the educators, two in graphic
design, one in journalism. Each educator had 3-4 years of university
Design Paerns for Data-Driven News Articles CHI ’24, May 11–16, 2024, Honolulu, HI, USA
teaching experience. Both online and oine workshops had a du-
ration of 2.5 hours each. We began with a 30-minute introduction
to data-driven journalism, data stories, and data visualization, then
introduced the ve article types and design patterns with examples.
After the introduction, participants were divided into groups of
two on a voluntary basis. In the online workshops, participants
joined in pairs, with one session featuring one participant. We
provided the Brieng article and the other materials. Each group was
asked to create one type of article chosen from Chart Descriptions
to In-Depth Investigations by using inspiration from the design
patterns. For oine workshops, we oered physical copies of the
design patterns and worksheets as cards. Each card contains the
title, icon, brief description and case of the design pattern. Online
workshops featured digital copies of those cards on a shared Miro
board. This creation phase took about 60-70 minutes. Participants
were told to think aloud and be free to discuss with their team
members.
Finally, we asked each group to present their article outline and
design rationale. The article outline mimics the structure of a data-
driven news article, with design pattern cards being selected and
dragged into the corresponding parts of the article. For example, se-
lecting a headline card and placing it at the beginning, accompanied
by a brief textual annotation explaining how the title was created
using the card prompt. Narrative cards are similarly selected and
placed in order, with a short textual description explaining the con-
tent. To reect on their design process and associated challenges,
we conducted semi-structured interviews with online workshop
participants and questionnaires with oine participants. Questions
centered around Whether our article types and design patterns were
helpful for creation, and if so, in what ways? What were the challenges
you encountered in your creation process? How can our approach be
improved? Did you employ other digital assistants, such as Chat-
GPT, during the creation process? If so, how did you utilize them? All
sessions of the online workshops were video and audio recorded.
Following the workshop, we conducted interviews with three ed-
ucators. We presented ve types of data-driven news articles and
related design patterns and engaged in semi-structured interviews
that started with two questions: Whether our approach could assist
learning and teaching, and if they would like to use these materials
for teaching, how they would implement them?
6.4 Results
The outcome of all workshops included 13 data-driven article out-
lines presented by assembled design pattern cards, brief textual
descriptions of their design rationales on worksheets, as well as
feedback gathered from interviews and questionnaires.
One group chose In-Depth Investigations on the topic of medi-
cal fees and insurance; they used ChatGPT for data queries. Eight
groups created Investigations, with the topic falling into two cat-
egories: regional comparisons and detailed analyses of particular
types of cancer, such as breast and lung cancer. Participants utilized
diverse sources for data collection, including datasets we supplied,
and enriched their research by querying additional data and ma-
terials through search engines, data from academic publications,
and ChatGPT. Four groups of students created Chart Descriptions,
with three groups highlighting the dierences in cancer prevalence
across regions and among dierent populations, and one group
focusing on the relationship between food temperature and the risk
of cancer.
Regarding the article construction process, we noted a variety
of approaches to structuring the articles. The design patterns were
used in a exible way rather than following the same workow.
Five groups focused on the data and visualization techniques rst,
then moved to the narrative section, while the other eight groups
discussed the narrative aspect rst, followed by the selection of the
visualization.
Classication of article types——Based on our observations
and feedback, we found that identifying the article type initially
helps to set expectations for the article’s content and guides the
scope of initial data searches. In our workshops, identifying and
determining the article type was the rst step in the construction
process, contributing to the overall planning of the article, and
forming an initial selection of the design pattern cards for the
framework. Group 8 participant 1 (G8P1): “The questions of what
type you plan to write made it easier for me to make choices on the
narrative; Educator 2 (Edu2): “Identifying the article type is an
important initial step in article construction, providing students with
a directional framework for their work.).
Design pattern usability——Our design patterns eectively
enable beginners to rapidly engage in hands-on practice, fa-
cilitating the creation of data-driven news articles. The type
of articles and design pattern cards gave participants a rich pool
of inspirational options for their writing. Even participants new
to data journalism and data visualization can discuss and use the
design pattern cards to frame the articles. Among the seven groups
who constructed Investigations, each group employed 9 to 12 narra-
tive patterns. The most frequently used data-driven text patterns
were Change over time (by 7 groups) and Contrasting pattern (by 6
groups). Additionally, participants made extensive use of context-
enhancing patterns, including Citing the authority and Summarize
(by all 7 groups), as well as Co-relation and causality and Domain
knowledge (by 6 groups). Participants’ pattern usage closely mir-
rored the patterns we coded in the Investigations, suggesting that
participants’ planning of the depth and direction of their inves-
tigative endeavors aligns with the established conventions of this
article type. G7P1: “The patterns are very comprehensive and work as
small molecules. Beginners like me were able to understand and use
them to build an article framework”; Edu1: “[design patterns] con-
tribute to the development of students’ abilities to organize material
eectively, which in turn encompasses both the richness of the content
and the clarity of its presentation”.
Due to the variation of article types constructed by participants,
it is challenging to identify clear trends in the use of dierent types
and design patterns. However, we still observed some design pat-
terns that aligned with participant preferences. For instance, the use
of headline patterns, Ask a Question (5 groups) and Sensationalism
(4 groups) are more frequently used, indicating that participants
aimed to create titles that eectively capture attention. Narrative
design patterns showed Contrast being frequently used (13 groups)
in data-driven narratives, followed by Change over time (7 groups).
In context-enhancing aspects, patterns like Summarize, Citing the
authority, and Correlation and causality were frequent (11 groups
each). Interestingly, Action was also widely used (9 groups).
CHI ’24, May 11–16, 2024, Honolulu, HI, USA Hao et al.
We also observed that 6 groups used ChatGPT during the cre-
ative process. They explored various directions for their inquiries,
including data sources (e.g., G1: ‘Could you suggest websites that
oer digital health reports containing relevant data on cancer?’),
understanding a topic (e.g., G2: ‘What are the factors that cause
cancer?’ ‘What foods are likely to cause cancer?’), participants used
design patterns to frame their questions enquiring ideas for specic
components (e.g., G1: ‘Provide a headline idea with a sensational
tone. and G2: ‘What are co-relation or causality between lifestyle
and cancer?’).
Providing diverse methods for presenting information broad-
ens the range of creative possibilities in writing. Integrating data
into articles and visualizations was a recurring challenge and focal
point in group discussions among the participants. They found our
design patterns useful in inspiring them to address this challenge.
For example, G4P2: After learning so many ways of interaction, I have
a clearer idea about how to deal with a large amount of data. The dis-
cussion in groups was mainly about making decisions among these
options such as visualization techniques and proper interactions to
be applied.
Increasing awareness of data transparency. Indicating the
sources of data is a demonstration of journalistic transparency [
64
].
The data source design patterns served as a reminder for partic-
ipants to emphasize data authenticity and credibility, facilitating
their inclusion of these elements in their articles. All of the partici-
pants used this pattern in their outlines.
Educators have expressed appreciation for our article types and
design patterns, believing that these materials can aid students
in their reporting-writing process. Regarding the usability of our
materials in teaching, Edu1, the lecturer in journalism, suggests
that our materials can be integrated into courses in news reporting
and interviewing in journalism. Journalism writing demands both
eciency and objective accuracy, which extends beyond sourc-
ing data or obtaining foundational materials. It also requires e-
cient textual narrative structures, an area where our materials can
provide valuable guidance. For teaching interviews in journalism,
“[narrative pattern cards] oer a variety of options, which can help
students organize the logical ow of their writing, thereby avoiding
major oversights or logical inconsistencies during the writing or inter-
view process”. Design instructor Edu2 expressed that our materials
can assist students in developing eective thinking patterns and
practical approaches. In the visualization techniques section, com-
monly used data visualization techniques are introduced, providing
students with valuable guidance for creating visualizations. He rec-
ommends, expanding the introduction of data visualization design
tools to provide further guidance for professionals specializing in data
visualization design.
7 DISCUSSION
7.1 Creating Data-Driven News Articles with
Design Patterns
By analyzing articles from data columns and the personal websites
of data journalists, we identify and describe a spectrum of ve
types of data-driven news articles, ranging from Quick Updates to
In-depth Investigations. Built on the existing classication of data
journalism by Google News Lab [
57
] and Nardelli [
55
], our article
types and design patterns systematically describe the corpus of
article structure, and components. This ready-to-use resource is
invaluable for teaching, designing, and engaging in discourse on
data-driven news articles.
Our design patterns oer several opportunities in the multi-
faceted process of creating data-driven news articles, which spans
data collection, analysis, visualization, and article composition.
Firstly, they act as cognitive aids for conceptualizing and structuring
data-driven articles, thus enhancing both writing and investiga-
tive skills. These patterns oer a rich information pool that helps
kickstart creative thinking and foster a logically structured ow. In-
spired by the successful use of cards for design thinking [
7
,
29
,
34
],
we create the multi-card format of design patterns to promote alter-
native methods of presentation and organization with digital and
physical activities, and stimulate fruitful discussions.
Additionally, design patterns can enhance the investigative qual-
ity of articles as they guide students in formulating questions and
in critically interpreting data (e.g., by Citing dierent viewpoints or
surface the process of Inference), as well as reduce the likelihood of
omitting crucial details (e.g., by Mentioning the data source). Essen-
tially, article type and design patterns oer a ready-to-use toolbox
for aspiring authors, beneting both educational and communica-
tive elds.
However, it is important to note that these patterns alone do not
guarantee best practices. They are meant to facilitate the article-
writing process by oering structured support. However, the even-
tual quality of the nal article depends on numerous factors, such
as writing style, data complexity, quality and suitability of visual-
izations, on the authoring side, as well as logical reasoning, data
literacy, and relevance, on the audience side. Future research could
focus on exploring how to involve this approach in current jour-
nalism education.
7.2 Eciency Gains and Formulaicity Risks
with Design Patterns for Data-Driven News
Article Creation
Although Section 5.2 presents some trends of patterns with specic
article types, it is crucial to remain cautious of the formalization
risks associated with rigidly applying these patterns. Even though
we counted, we opted not to disclose the frequency of pattern usage
by article type to avoid creating potential biases for participants.
Presenting such statistics from the sample could inadvertently es-
tablish a perceived ‘standard’ for how certain patterns should be
used for specic article types. This could lead to a ‘one-size-ts-all’
approach, stiing individualized or innovative design decisions tai-
lored to unique cases and communication goals. In the workshop,
we did not observe the xed modes of combination patterns. Partic-
ipants presented various creative approaches. Our design patterns
served as a toolbox rather than a standard structure. Participants
actively discussed their choices of patterns with their collaborators
in the creation process.
Balancing creative freedom with structured guidance is, however,
a trade-o in support tools for article creation[
52
]. Future research
could delve into the relationships between article types, design
Design Paerns for Data-Driven News Articles CHI ’24, May 11–16, 2024, Honolulu, HI, USA
patterns, and communication objectives with a larger sample, po-
tentially informing the development of design tools that oer more
nuanced guidance.
7.3 Adaptions and Extensions
Our design patterns for data-driven news articles can extend be-
yond traditional formats to inform the design of other storytelling
genres, such as scrollytelling or slideshow articles that also incor-
porate titles, textual descriptions, visualizations, and references to
data sources. Additionally, the design patterns could be adapted
to various social media platforms where long-form investigative
pieces are distilled into concise teasers to quickly inform and attract
specic audiences, including through visually engaging formats
such as GIFs [
62
]. Thus, our design patterns could provide the basis
for the development of a exible framework for content adaptation
across platforms that could inform the development of targeted,
platform-specic information algorithms.
Furthermore, our design patterns can be combined by using
other low-level narrative devices that serve specic intents, such as
the 6 detailed time-oriented narrative sequences (i.e., Chronology,
Trace-back, Trailer, Recurrence, Halfway-back, and Anchor) [
40
],
or the 18 narrative design patterns for data-driven storytelling [
6
]
such as Convention breaking and Concretize to enrich the narration
in news reporting. By adapting to other layouts, our design patterns
can be used to create data-driven news articles in other forms of
presentation such as posters, data comics, or data videos.
Finally, our design patterns could help human-machine collab-
orations. We noted that workshop participants utilized various
forms of assistance during the article creation process, includ-
ing using ChatGPT to seek inspiration for diverse facets of the
topic and to locate data resources. Within the realm of journal-
ism, technologies can be seen as collaborative human partners,
or ‘co-creators of journalism. [
51
]. However, current automated
technologies are limited in interpreting cultural sensitivity or hu-
man emotions which are essential factors for investigative types
of journalistic reporting [
70
]. There is indeed a gap between au-
thoring tools, AI-supported tools [
44
], and AI-generator tools [
54
].
Recommended approaches for the latter are typically solely based
on data features or design guidelines ignoring creators’ intent [
13
].
Hence, signicant untapped potential exists for tools that bridge
technologies with creative practice. Our design patterns could serve
as a modular framework to facilitate more eective interactions
between creators and AI-based generation tools.
7.4 Limitations of the Work
We acknowledge that our article samples are primarily from West-
ern English-speaking media. Expanding the dataset to articles in
dierent languages and from dierent cultures could provide valu-
able cultural and complementary perspectives. While our sampled
corpus showcases a degree of diversity, limitations in sample size
and representativeness remain; e.g., our samples represent what
Sparks [
63
] describes as the ‘serious press’ and ‘semi-serious press’,
characterized by a blend of traditional news and soft news features.
However, our study did not include much of the ‘serious-popular
press’ and ‘news stand tabloid press’, known for their visual em-
phasis and focus on scandal, sports, and entertainment, and the
ones which leans heavily towards scandal and entertainment while
retaining some serious news elements. One could think that our
sampling initiated with articles on COVID-19 could be a further
limitation. However, when we compared design patterns across
articles on COVID-19 with those on other topics, we found no
signicant dierences in their frequency and application.
Eventually, while our paper focuses on the structure and com-
ponents of data-driven news articles, other elements like layout,
font, color, and presentation devices warrant further investigation.
While article types and design patterns serve as useful aids for con-
tent creation, eective teaching and learning in visualization [
5
]
and data-driven news should also incorporate other critical com-
petencies, such as data and visualization literacy as well as ethical
approaches to data and journalistic practice. We focus on novice cre-
ators planning to create data-driven news articles. Future research
could engage with professionals to expand or rene the patterns
and usage.
8 CONCLUSION
In this paper, we describe design patterns for data-driven news
articles to support literacy and prociency in writing data-driven
news articles. Building on existing classications in data journalism
and analyzing contemporary examples, we also identied ve types
of data-driven news articles tailored for various scenarios, aiding
creators in setting expectations and swiftly constructing a frame-
work for their articles. Subsequently, we approach the article from a
holistic perspective, identifying design patterns under eleven com-
mon components of data-driven news articles: headline, news lead,
data source, narrative, image, visualization title, visualization tech-
niques, visualization annotation, caption, interaction, and video.
These design patterns were rened through expert interviews and
workshops. Results suggest that our design patterns could guide
beginners in quickly getting started and improving their eciency
in constructing data-driven news articles.
REFERENCES
[1]
2023. Reuters Institute and the University of Oxford Issues Report Entitled
’Reuters Institute Digital News Report 2022’.
[2]
R. Amar and J. Stasko. 2004. BEST PAPER: A Knowledge Task-Based Framework
for Design and Evaluation of Information Visualizations. In IEEE Symposium on
Information Visualization. 143–150. https://doi.org/10.1109/INFVIS.2004.10
[3]
Faheem Aslam, Tahir Mumtaz Awan, Jabir Hussain Syed, Aisha Kashif, and
Mahwish Parveen. 2020. Sentiments and emotions evoked by news headlines of
coronavirus disease (COVID-19) outbreak. Humanities & social sciences commu-
nications 7, 1 (2020), 1–9.
[4]
Benjamin Bach, Euan Freeman, Ale Abdul-Rahman, Cagatay Turkay, Saiful
Khan, Yulei Fan, and Min Chen. 2023. Dashboard Design Patterns. IEEE
Transactions on Visualization and Computer Graphics 29, 1 (2023), 342–352.
https://doi.org/10.1109/TVCG.2022.3209448
[5]
Benjamin Bach, Mandy Keck, Fateme Rajabiyazdi, Tatiana Losev, Isabel Meirelles,
Jason Dykes, Robert S Laramee, Mashael AlKadi, Christina Stoiber, Samuel Huron,
et al. 2023. Challenges and opportunities in data visualization education: A call
to action. IEEE Transactions on visualization and computer graphics (2023).
[6]
Benjamin Bach, Moritz Stefaner, Jeremy Boy, Steven Drucker, Lyn Bartram, Jo
Wood, Paolo Ciuccarelli, Yuri Engelhardt, Ulrike Koeppen, and Barbara Tversky.
2018. Narrative design patterns for data-driven storytelling. In Data-driven
storytelling. AK Peters/CRC Press, 107–133.
[7]
Benjamin Bach, Zezhong Wang, Matteo Farinella, Dave Murray-Rust, and
Nathalie Henry Riche. 2018. Design Patterns for Data Comics. In Proceedings of
the 2018 CHI Conference on Human Factors in Computing Systems (Montreal QC,
Canada) (CHI ’18). Association for Computing Machinery, New York, NY, USA,
1–12. https://doi.org/10.1145/3173574.3173612
[8]
Stefano Bocconi, Frank Nack, and Lynda Hardman. 2005. Supporting the Gen-
eration of Argument Structure within Video Sequences. In Proceedings of the
CHI ’24, May 11–16, 2024, Honolulu, HI, USA Hao et al.
Sixteenth ACM Conference on Hypertext and Hypermedia (Salzburg, Austria) (HY-
PERTEXT ’05). Association for Computing Machinery, New York, NY, USA, 75–84.
https://doi.org/10.1145/1083356.1083371
[9]
Eddy Borges-Rey. 2016. Unravelling Data Journalism: A study of data journalism
practice in British newsrooms. Journalism practice 10, 7 (2016), 833–843.
[10]
Liliana Bounegru, Liliana Bounegru, and Jonathan Grey. 2021. The data journal-
ism handbook : towards a critical data practice / edited by Liliana Bounegru and
Jonathan Gray. Amsterdam University Press, Amsterdam.
[11]
Jeremy Boy, Francoise Detienne, and Jean-Daniel Fekete. 2015. Storytelling in
Information Visualizations: Does It Engage Users to Explore Data?. In Proceedings
of the 33rd Annual ACM Conference on Human Factors in Computing Systems
(Seoul, Republic of Korea) (CHI ’15). Association for Computing Machinery, New
York, NY, USA, 1449–1458. https://doi.org/10.1145/2702123.2702452
[12]
Paul Bradshaw. 2011. The inverted pyramid of data journalism. Online Journalism
Blog 7 (2011).
[13]
Qing Chen, Shixiong Cao, Jiazhe Wang, and Nan Cao. 2023. How Does Au-
tomation Shape the Process of Narrative Visualization: A Survey of Tools.
IEEE Transactions on Visualization and Computer Graphics (2023), 1–20. https:
//doi.org/10.1109/TVCG.2023.3261320
[14]
Qing Chen, Ziyan Liu, Chengwei Wang, Xingyu Lan, Ying Chen, Siming Chen,
and Nan Cao. 2022. VizBelle: A Design Space of Embellishments for Data Visual-
ization. arXiv.org (2022).
[15]
Siming Chen, Jie Li, Gennady Andrienko, Natalia Andrienko, Yun Wang, Phong H.
Nguyen, and Cagatay Turkay. 2020. Supporting Story Synthesis: Bridging the Gap
between Visual Analytics and Storytelling. IEEE Transactions on Visualization
and Computer Graphics 26, 7 (2020), 2499–2516. https://doi.org/10.1109/TVCG.
2018.2889054
[16]
Fanny Chevalier, Melanie Tory, Bongshin Lee, Jarke van Wijk, Giuseppe Santucci,
Marian Dörk, and Jessica Hullman. 2018. From analysis to communication:
Supporting the lifecycle of a story. In Data-Driven Storytelling. AK Peters/CRC
Press, 151–183.
[17]
Celeste M Condit, Alex Ferguson, Rachel Kassel, Chitra Thadhani, Holly Cather-
ine Gooding, and Roxanne Parrott. 2001. An exploratory study of the impact
of news headlines on genetic determinism. Science Communication 22, 4 (2001),
379–395.
[18]
Zeyu Dai, Himanshu Taneja, and Ruihong Huang. 2018. Fine-grained structure-
based news genre categorization. In Proceedings of the Workshop Events and Stories
in the News 2018. 61–67.
[19]
Ensheng Dong, Hongru Du, and Lauren Gardner. 2020. An interactive web-based
dashboard to track COVID-19 in real time. The Lancet infectious diseases 20, 5
(2020), 533–534.
[20]
Brent Dykes. 2019. Eective Data Storytelling - How to Drive Change with Data,
Narrative, and Visuals. John Wiley & Sons, Newark.
[21]
The Economist. 2021. The pandemic’s true death toll. https://www.economist.
com/graphic-detail/coronavirus-excess-deaths-estimates. [last accessed: Mar 2,
2023].
[22]
Gerald Frank Else. 1938. Aristotle on the Beauty of Tragedy. Harvard studies in
classical philology 49 (1938), 179–204.
[23]
James S Ettema and Theodore L Glasser. 1998. Custodians of conscience: Inves-
tigative journalism and public virtue. Columbia University Press.
[24]
Gustav Freytag and Elias J. MacEwan. 1895. Freytag’s Technique of the drama :
an exposition of dramatic composition and art. Benjamin Bloom, New York.
[25]
D. Fry. 2009. Stack of Blocks. https://donfry.wordpress.com/2009/06/14/stack-
of-blocks/
[26]
Lucas Galan, Jordan Osserman, Tim Parker, and Matthew Taylor. 2019. How
young people consume news and the implications for mainstream media. (2019).
[27]
Jonathan Gray, Lucy Chambers, and Liliana Bounegru. 2012. The Data Journalism
Handbook. O’Reilly Media, Incorporated, Piraí.
[28]
The Guardian. [n. d.]. One million coronavirus deaths: how did we get
here? https://www.theguardian.com/world/ng-interactive/2020/sep/29/one-
million-coronavirus-deaths-how-did-we-get-here-covid. [last accessed: Nov 26,
2023].
[29]
Kim Halskov and Peter Dalsgård. 2006. Inspiration card workshops. In Proceedings
of the 6th conference on Designing Interactive systems. 2–11.
[30]
Michael Hameleers. 2019. Partisan Media, Polarized Audiences? A Qualitative
Analysis of Online Political News and Responses in the United States, U.K., and
The Netherlands. International journal of public opinion research 31, 3 (2019),
485–505.
[31]
James T Hamilton. 2016. Democracy’s detectives: The economics of investigative
journalism. Harvard University Press.
[32]
Timothy Harrower. 2006. Inside Reporting: A Practical Guide to the Craft of
Journalism. https://api.semanticscholar.org/CorpusID:111049843
[33]
Bahareh Heravi. 2022. Storytelling Structures in Data Journalism: Introducing
the Water Tower structure. In Computation+ Journalism 2022.
[34]
Eva Hornecker. 2010. Creative Idea Exploration within the Structure of a Guiding
Framework: The Card Brainstorming Game. In Proceedings of the Fourth Interna-
tional Conference on Tangible, Embedded, and Embodied Interaction (Cambridge,
Massachusetts, USA) (TEI ’10). Association for Computing Machinery, New York,
NY, USA, 101–108. https://doi.org/10.1145/1709886.1709905
[35]
Jessica Hullman, Steven Drucker, Nathalie Henry Riche, Bongshin Lee, Danyel
Fisher, and Eytan Adar. 2013. A Deeper Understanding of Sequence in Narrative
Visualization. IEEE Transactions on Visualization and Computer Graphics 19, 12
(2013), 2406–2415. https://doi.org/10.1109/TVCG.2013.119
[36]
Waqas Javed and Niklas Elmqvist. 2012. Exploring the design space of composite
visualization. In 2012 IEEE Pacic Visualization Symposium. 1–8. https://doi.org/
10.1109/PacicVis.2012.6183556
[37]
Precision Journalism. 1973. A Reporter’s Introduction to Social Science Methods.
[38]
Nicolas Kayser-Bril, Anastasia Valeeva, and Irina Radchenko. 2016. Trans-
formation of communication processes: Data journalism. arXiv preprint
arXiv:1605.01956 (2016).
[39]
Robert Kosara and Jock Mackinlay. 2013. Storytelling: The Next Step for Visual-
ization. Computer 46, 5 (2013), 44–50. https://doi.org/10.1109/MC.2013.36
[40]
Xingyu Lan, Xinyue Xu, and Nan Cao. 2021. Understanding Narrative Linearity
for Telling Expressive Time-Oriented Stories. In Proceedings of the 2021 CHI
Conference on Human Factors in Computing Systems (Yokohama, Japan) (CHI ’21).
Association for Computing Machinery, New York, NY, USA, Article 604, 13 pages.
https://doi.org/10.1145/3411764.3445344
[41]
Shahid Latif, Siming Chen, and Fabian Beck. 2021. A Deeper Understanding
of Visualization-Text Interplay in Geographic Data-driven Stories. In Computer
Graphics Forum, Vol. 40. Wiley Online Library, 311–322.
[42]
Bongshin Lee, Nathalie Henry Riche, Petra Isenberg, and Sheelagh Carpendale.
2015. More Than Telling a Story: Transforming Data into Visually Shared Stories.
IEEE Computer Graphics and Applications 35, 5 (2015), 84–90. https://doi.org/10.
1109/MCG.2015.99
[43]
Wiebke Loosen, Julius Reimer, and Fenja De Silva-Schmidt. 2020. Data-driven
reporting: An on-going (r) evolution? An analysis of projects nominated for the
Data Journalism Awards 2013–2016. Journalism 21, 9 (2020), 1246–1263.
[44]
Min Lu, Chufeng Wang, Joel Lanir, Nanxuan Zhao, Hanspeter Pster, Daniel
Cohen-Or, and Hui Huang. 2020. Exploring Visual Information Flows in Info-
graphics. In Proceedings of the 2020 CHI Conference on Human Factors in Computing
Systems (Honolulu, HI, USA) (CHI ’20). Association for Computing Machinery,
New York, NY, USA, 1–12. https://doi.org/10.1145/3313831.3376263
[45]
Alan Lundgard and Arvind Satyanarayan. 2022. Accessible Visualization via
Natural Language Descriptions: A Four-Level Model of Semantic Content. IEEE
Transactions on Visualization and Computer Graphics 28, 1 (2022), 1073–1083.
https://doi.org/10.1109/TVCG.2021.3114770
[46] Anna McKane. 2013. News writing. News Writing (2013), 1–192.
[47]
S. McKenna, N. Henry Riche, B. Lee, J. Boy, and M. Meyer. 2017. Visual Narrative
Flow: Exploring Factors Shaping Data Visualization Story Reading Experiences.
Computer graphics forum 36, 3, 377–387.
[48]
Faten El Outa, Patrick Marcel, Veronika Peralta, and Panos Vassiliadis. 2023.
Highlighting the Importance of Intentional Aspects in Data Narrative Crafting
Processes. Information Systems Frontiers (2023), 1–17.
[49]
Anna Podara, Maria Matsiola, Theodora A Maniou, and George Kalliris. 2019.
News usage patterns of young adults in the era of Interactive Journalism. Strategy
& Development Review 9 (2019).
[50]
Horst Pöttker. 2003. News and its communicative quality. The «inverted pyra-
mid»—Origin and establishment of a professional standard in journalism: Die
«Umgekehrte Pyramide»—Ursprung und Durchsetzung eines journalistischen
Standards. Publizistik 48 (2003), 414–426.
[51]
Alex Primo and Gabriela Zago. 2015. Who And What Do Journalism?: An actor-
network perspective. Digital journalism 3, 1 (2015), 38–52.
[52]
Larissa Pschetz, Koji Yatani, and Darren Edge. 2014. TurningPoint: narrative-
driven presentation planning. In Proceedings of the SIGCHI Conference on Human
Factors in Computing Systems. 1591–1594.
[53]
Lea Püchel and Christian-Mathias Wellbrock. 2021. Examining the digital re-
newal of news communication: A categorization of presentation modes in digital
journalism. New Media & Society (2021), 14614448211059488.
[54]
Chunyao Qian, Shizhao Sun, Weiwei Cui, Jian-Guang Lou, Haidong Zhang, and
Dongmei Zhang. 2021. Retrieve-Then-Adapt: Example-based Automatic Genera-
tion for Proportion-related Infographics. IEEE Transactions on Visualization and
Computer Graphics 27, 2 (2021), 443–452. https://doi.org/10.1109/TVCG.2020.
3030448
[55]
Alastair Reid. 2015. ’Without humanity, data alone is meaningless’: Data
journalism insights from the Guardian. https://www.journalism.co.uk/news/-
without-humanity-data-alone-is-meaningless-data-journalism-insights-from-
the-guardian/s2/a564652/. [last accessed: Mar 2, 2023].
[56]
Donghao Ren, Matthew Brehmer, Bongshin Lee, Tobias Höllerer, and Eun Kyoung
Choe. 2017. ChartAccent: Annotation for data-driven storytelling. In 2017 IEEE
Pacic Visualization Symposium (PacicVis). 230–239. https://doi.org/10.1109/
PACIFICVIS.2017.8031599
[57]
Simon Rogers, Jonathan Schwabish, and Danielle Bowers. 2017. Data journalism
in 2017: The current state and challenges facing the eld today. Google News Lab
(2017).
Design Paerns for Data-Driven News Articles CHI ’24, May 11–16, 2024, Honolulu, HI, USA
[58]
Michael Schudson. 2014. Origins of the freedom of information act in the United
States. Transparency in Politics and the Media. Accountability and Open Gov-
ernment. Reuters Institute for the Study of Journalism & IB Tauris, Oxford (2014),
1–18.
[59]
Edward Segel and Jerey Heer. 2010. Narrative Visualization: Telling Stories
with Data. IEEE Transactions on Visualization and Computer Graphics 16, 6 (2010),
1139–1148. https://doi.org/10.1109/TVCG.2010.179
[60]
Yang Shi, Xingyu Lan, Jingwen Li, Zhaorui Li, and Nan Cao. 2021. Communicating
with Motion: A Design Space for Animated Visual Narratives in Data Videos. In
Proceedings of the 2021 CHI Conference on Human Factors in Computing Systems
(Yokohama, Japan) (CHI ’21). Association for Computing Machinery, New York,
NY, USA, Article 605, 13 pages. https://doi.org/10.1145/3411764.3445337
[61]
Dilruba Showkat and Eric P. S. Baumer. 2021. Where Do Stories Come From?
Examining the Exploration Process in Investigative Data Journalism. Proc. ACM
Hum.-Comput. Interact. 5, CSCW2, Article 390 (oct 2021), 31 pages. https://doi.
org/10.1145/3479534
[62]
Xinhuan Shu, Aoyu Wu, Junxiu Tang, Benjamin Bach, Yingcai Wu, and Huamin
Qu. 2020. What makes a data-gif understandable? IEEE Transactions on Visual-
ization and Computer Graphics 27, 2 (2020), 1492–1502.
[63]
Colin Sparks and John Tulloch. 2000. Tabloid tales: Global debates over media
standards. Rowman & Littleeld.
[64]
Florian Stalph. 2018. Classifying Data Journalism: A content analysis of daily
data-driven stories. Journalism practice 12, 10 (2018), 1332–1350.
[65]
Nicole Sultanum, Fanny Chevalier, Zoya Bylinskii, and Zhicheng Liu. 2021. Lever-
aging Text-Chart Links to Support Authoring of Data-Driven Articles with
VizFlow. In Proceedings of the 2021 CHI Conference on Human Factors in Computing
Systems (Yokohama, Japan) (CHI ’21). Association for Computing Machinery, New
York, NY, USA, Article 16, 17 pages. https://doi.org/10.1145/3411764.3445354
[66]
Financial Times. 2021. Charts that work: FT visual vocabulary guide. https:
//www.ft.com/content/c7bb24c9-964d-479f-ba24-03a2b2df6e85. [last accessed:
Aug 22, 2023].
[67]
Turo I Uskali and Heikki Kuutti. 2015. Models and streams of data journalism.
The journal of media innovations 2, 1 (2015), 77–88.
[68]
Andreas Veglis and Charalampos Bratsas. 2017. Towards a taxonomy of data
journalism. Journal of Media Critiques 3, 11 (2017), 109–121.
[69]
Wibke Weber, Martin Engebretsen, and Helen Kennedy. 2018. Data stories:
Rethinking journalistic storytelling in the context of data journalism. Studies in
communication sciences 2018, 1 (2018), 191–206.
[70]
Shangyuan Wu, Edson C Tandoc Jr, and Charles T Salmon. 2019. Journalism
Recongured: Assessing Human–machine Relations and the Autonomous Power
of Automation in News Production. Journalism studies 20, 10 (2019), 1440–1457.
[71]
Leni Yang, Xian Xu, XingYu Lan, Ziyan Liu, Shunan Guo, Yang Shi, Huamin Qu,
and Nan Cao. 2022. A Design Space for Applying the Freytag’s Pyramid Structure
to Data Stories. IEEE Transactions on Visualization and Computer Graphics 28, 1
(2022), 922–932. https://doi.org/10.1109/TVCG.2021.3114774
[72]
Ji Soo Yi, Youn ah Kang, John Stasko, and J.A. Jacko. 2007. Toward a Deeper
Understanding of the Role of Interaction in Information Visualization. IEEE
Transactions on Visualization and Computer Graphics 13, 6 (2007), 1224–1231.
https://doi.org/10.1109/TVCG.2007.70515
[73]
Ogan M Yigitbasioglu and Oana Velcu. 2012. A review of dashboards in perfor-
mance management: Implications for design and research. International Journal
of Accounting Information Systems 13, 1 (2012), 41–59.
[74]
Mary Lynn Young, Alfred Hermida, and Johanna Fulda. 2018. What makes for
great data journalism? A content analysis of data journalism awards nalists
2012–2015. Journalism practice 12, 1 (2018), 115–135.
[75]
Yixuan Zhang, Yifan Sun, Lace Padilla, Sumit Barua, Enrico Bertini, and Andrea G
Parker. 2021. Mapping the Landscape of COVID-19 Crisis Visualizations. In
Proceedings of the 2021 CHI Conference on Human Factors in Computing Systems
(Yokohama, Japan) (CHI ’21). Association for Computing Machinery, New York,
NY, USA, Article 608, 23 pages. https://doi.org/10.1145/3411764.3445381
CATALOG OF DATA-DRIVEN NEWS ARTICLE
SAMPLES
[G1]
The Guardian. Elections 2022: live council results for England, Scotland and Wales. 2022.
https://www.theguardian.com/politics/ng-interactive/2022/may/05/elections-2022- results-
live-local-council-england-scotland- wales
[G2]
The Guardian. Truss or Sunak? Who do bookies favour to be the next PM? . 2022. https://
www.theguardian.com/politics/2022/jul/27/tory-leadership-latest-odds-tracker-bookies-
next-pm-sunak-truss
[G3]
The Guardian. How did your MP vote on the fracking motion? 2017. https://www.theguardian.
com/environment/ng-interactive/2022/oct/19/how-did-your-mp-vote-on- the-fracking-motion
[G4]
The Guardian. Covid in the UK cases, deaths and vaccinations the latest numbers. 2022.
https://www.theguardian.com/world/2022/jun/01/covid-19-uk-cases-deaths- vaccinations-
latest-numbers-coronavirus-national-data
[G5]
The Guardian. Liz Truss v Rishi Sunak a timeline of their roads to Tory nal two. 2022. https:
//www.theguardian.com/politics/ng-interactive/2022/aug/15/liz-truss-v-rishi-sunak-a-timeline-
of-their-roads-to-tory-nal-two
[G6]
The Guardian. The worldwide scale of the Uber les in numbers. 2022. https://www.
theguardian.com/news/2022/jul/15/the-worldwide-scale-of-the-uber-les-in-numbers
[G7]
The Guardian. Europe’s record summer of heat and res visualised. 2022. https://www.
theguardian.com/environment/ng-interactive/2022/jul/26/how-europe-has-been-hit-by-
record-re-damage-and-temperatures
[G8]
The Guardian. Boris Johnson’s tumultuous three years as prime minister in charts. 2022.
https://www.theguardian.com/politics/2022/jul/07/boris-johnsons-tumultuous-three- years-
as-prime-minister-in-charts
[G9]
The Guardian. Are enough people vaccinated in time for England’s ‘freedom day’? 2021.
https://www.theguardian.com/world/2021/jul/19/are-enough-people-vaccinated-in-time-
for-englands-freedom-day
[G10]
The Guardian. Covid vaccine gures lay bare global inequality as global target missed. 2021.
https://www.theguardian.com/global-development/2022/jul/21/covid-vaccine-gures-lay-
bare-global-inequality-as-global- target-missed
[G11]
The Guardian. In numbers what is fuelling Britain’s cost of living crisis. 2022. https:
//www.theguardian.com/business/2022/feb/03/in-numbers-britains-cost-of-living-crisis
[G12]
The Guardian. Which countries are the most (and least) committed to reducing inequal-
ity. 2017. https://www.theguardian.com/inequality/datablog/2017/jul/17/which-countries-
most-and-least-committed-to-reducing-inequality-oxfam-d
[G13]
The Guardian. Why are some areas of the UK in drought A visual guide. 2022. https:
//www.theguardian.com/environment/2022/aug/19/why-areas-uk-drought-visual-guide
[G14]
The Guardian. More than 25m drink from the worst US water systems, with Latinos
most exposed. 2021. https://www.theguardian.com/us-news/2021/feb/26/worst-us-water-
systems-latinos-most-exposed
[Ft1]
Financial Times. Tory leadership election tracker: Rishi Sunak and Liz Truss remain in
race to be PM. 2022. https://www.ft.com/content/5ed407bd-ac0d-4293-a0fe-c01dfa15d9d2
[Ft2]
Financial Times. Covid-19 vaccine tracker: the global race to vaccinate. 2022. https://ig.ft.
com/coronavirus-vaccine-tracker/?areas=gbr&areas=isr&areas=usa&areas=eue&areas=are&
areas=chn&areas=chl&cumulative=1&doses=total&populationAdjusted=1
[Ft3]
Financial Times. Biden vs Trump:live result 2020. 2020. https://ig.ft.com/us-election-2020/
[Ft4]
Financial Times. High UK ination hits sales, condence and activity. 2022. https://www.ft.
com/content/ec8c3bfb-5473-4776-b483-f0a531940c46
[Ft5]
Financial Times. How the coronavirus pandemic has disrupted the US Democratic primary
calendar. 2020. https://www.ft.com/content/f3bb0944-4437-11ea-abea-0c7a29cd66fe
[Ft6]
Financial Times. UK general election poll tracker. 2019.https://www.ft.com/content/263615ca-
d873-11e9-8f9b-77216ebe1f17
[Ft7]
Financial Times. Climate graphic of the week Arctic warming four times faster than rest
of the planet. 2022.https://www.ft.com/content/9bb32c6f-b3dc-46bd-944e-9e583407b866
[Ft8]
Financial Times. Older population in England and Wales hits record high. 2022.https:
//www.ft.com/content/c8334a2a-cef9-4ed8-9470-5a4dbcfcd351
[Ft9]
Financial Times. Carmakers’ battery plans in peril as raw material costs soar. 2022.https:
//www.ft.com/content/b4002e49-07ce-41d8- 9d3b-b6ed55af798c
[Ft10]
Financial Times. Lab-grown meat maker Eat Just unable to capitalise on Malaysia chicken
ban. 2022.https://www.ft.com/content/f986f084-9bb7-4f3a- a91b-1a5c783fc257
[Ft11]
Financial Times. UK public nances on ‘unsustainable path’, says OBR. 2022.https://www.
ft.com/content/ae63f4fb- 184e-42b0-8d46-7a8a9be89cf4
[Ft12]
Financial Times. First round of 2022 French election in charts. 2022.https://www.ft.com/
content/2dd79519-c478-4660-886e-316c5cb553b3
[Ft13]
Financial Times. Global ination tracker see how your country compares on rising prices.
2023.https://www.ft.com/content/088d3368-bb8b-43- 9df7-a7680d4d81b2
[Ft14]
Financial Times. Can the EU wean itself o Russian gas. 2022.https://ig.ft.com/europes-
race-to-replace-russian-gas/
[Ft15]
Financial Times. The Queen’s 70-year reign in 10 charts. 2022.https://www.ft.com/content/
4ef98f0a-47c2-47e9-b93a-7e39f1d2a4d7
[Ft16]
Financial Times. Would carbon food labels change the way you shop? 2022.https://ig.ft.
com/carbon-food-labelling/
[Ft17]
Financial Times. Why ‘the worker’s market’ is not delivering for Americans? 2022.https:
//www.ft.com/content/5a13b24b-a16c-429c- 9128-d48dd183be5c
[Ft18]
Financial Times. Where are all the workers? The US states and sectors with the tightest
jobs markets. 2022.https://www.ft.com/content/9c5b5025-a995-494c- 83a5-65ef04952a1a
[Ft19]
Financial Times. Baby bust: economic stimulus helps births rebound from pandemic.
2022.https://www.ft.com/content/32436917-00b8-447d- 8d6c-41f4be72b03f
[Ft20]
Financial Times. The global stagation shock of 2022: how bad could it get? 2022.https:
//www.ft.com/content/d490ef4e-3187-471e- 84-9c065871a1a5
[B1]
BBC News. Covid map: Coronavirus cases, deaths, vaccinations by country. 2022.https:
//www.bbc.co.uk/news/world-51235105
[B2]
BBC News. What is the energy price cap and how high will bills go? 2023.https://www.
bbc.com/news/business-58090533
[B3]
BBC News. Brexit deal: How did my MP vote on the Withdrawal Agreement Bill? 2019.https:
//www.bbc.co.uk/news/uk-politics-50145265
[B4]
BBC News. Coronavirus: What is the R number and how is it calculated? 2021.https:
//www.bbc.co.uk/news/health-52473523
[T1]
The Times. World Cup 2022 Results, xtures, stats and tables. 2022.https://www.thetimes.
co.uk/sport/football/world-cup/xtures-results
[T2]
The Times. Don’t Trust Your Gut by Seth Stephens-Davidowitz can data make us
happy? 2022.https://www.thetimes.co.uk/article/dont-trust-your-gut-by-seth-stephens-
davidowitz-can-data-make-us-happy-kfjm9926l
[T3]
The Times. Digital banks beat the high street on customer service. 2022.https://www.
thetimes.co.uk/article/online-banks-outrank-high-street- rivals-for-customer-service-x6jwsxt5r
[T4]
The Times. Hospital treatment is still a waiting game. 2022.https://www.thetimes.co.uk/
article/hospital-treatment-is-still-a-waiting-game-pk5d85bfz
[T5]
The Times. The ve charts that explain the energy crisis. 2022.https://www.thetimes.co.
uk/article/the-ve-charts-that-explain-the-energy-crisis-lp8h3gcs6
[T6]
The Times. Tory members are making the decision - but do they represent public opinion.
2022.https://www.thetimes.co.uk/article/tory-members-are-making-the-decision-but-do-
they-represent-public-opinion-m80jnscqq
[T7]
The Times. How much did the Platinum Jubilee cost and who is paying. 2022.https://www.
thetimes.co.uk/article/how-much-did-the-platinum-jubilee-cost-and-who- is-paying-ww28ptqbf
CHI ’24, May 11–16, 2024, Honolulu, HI, USA Hao et al.
[T8]
The Times. What is the UK’s true Covid death toll. 2022.https://www.thetimes.co.uk/article/
what-is-the-uks-true- covid-death-toll- lzpb0xnqs
[E1]
The Economist. Does Gen Z spend too much time on social media? 2022.https://www.
economist.com/graphic-detail/2022/08/10/does-gen- z-spend-too-much-time- on-social-media
[E2]
The Economist. Business. 2022.https://www.economist.com/the-world-this-week/2022/08/
18/business
[E3]
The Economist. Business. 2022.https://www.economist.com/the-world-this-week/2022/07/
14/business
[E4]
The Economist. Global living standards are moving in the wrong direction. 2022.https://
www.economist.com/graphic-detail/2022/09/08/global-living-standards-are-moving-in-the-
wrong-direction
[E5]
The Economist. Elizabeth II was the longest-reigning monarch in British history. 2022.https:
//www.economist.com/graphic-detail/2022/09/08/elizabeth-ii-was-the-longest-reigning-
monarch-in-british-history
[E6]
The Economist. The world is almost back to pre-covid activity levels. 2022.https://www.
economist.com/graphic-detail/2022/09/08/the-world-is-almost-back-to-pre-covid-activity-
levels
[E7]
The Economist. Another mass shooting in America. 2022.https://www.economist.com/
graphic-detail/2022/07/05/another-mass-shooting-in-america
[E8]
The Economist. Surging fossil-fuel emissions are ruining carbon dating. 2022.https://www.
economist.com/graphic-detail/2022/08/12/surging-fossil-fuel-emissions-are-ruining-carbon-
dating
[E9]
The Economist. Better measurement would help reduce water consumption. 2022.https://
www.economist.com/britain/2022/08/11/better-measurement-would-help-reduce-water-
consumption
[E10]
The Economist. The Big Mac index. 2022.https://www.economist.com/big-mac-index?
fsrc=core-app-economist
[E11]
The Economist. Friendship across class lines may boost social mobility and decrease
poverty. 2022.https://www.economist.com/graphic-detail/2022/08/11/friendship-across-
class-lines-may-boost-social-mobility-and-decrease-poverty
[E12]
The Economist. Tracking covid-19 excess deaths across countries. 2022.https://www.
economist.com/graphic-detail/coronavirus-excess-deaths-tracker
[E13]
The Economist. What Spotify data show about the decline of English. 2022.https://www.
economist.com/interactive/graphic-detail/2022/01/29/what-spotify-data-show-about-the-
decline-of- english
[E13]
The Economist. What Spotify data show about the decline of English. 2022.https://www.
economist.com/interactive/graphic-detail/2022/01/29/what-spotify-data-show-about-the-
decline-of- english
[E14]
The Economist. There have been 7m-13m excess deaths worldwide during the pandemic.
2021.https://www.economist.com/brieng/2021/05/15/there-have-been-7m- 13m-excess-deaths-
worldwide-during-the-pandemic
[E15]
The Economist. How we estimated the true death toll of the pandemic. 2021.https://www.
economist.com/graphic-detail/2021/05/13/how-we-estimated-the-true-death-toll-of- the-
pandemic
[E16]
The Economist. Why it is too early to say the world economy is in recession. 2022.https://
www.economist.com/nance-and-economics/2022/07/24/why-it-is-too-early-to-say-the-
world-economy-is-in-recession
[NYT1]
The New York Times. Coronavirus in the U.S.: Latest Map and Case Count. 2021.https:
//www.nytimes.com/interactive/2021/us/covid-cases.html
[NYT2]
The New York Times. Tracking Dangerous Heat in the U.S.. 2022.https://www.nytimes.
com/interactive/2022/us/heat-wave-map-tracker.html
[NYT3]
The New York Times. Path to 218 Tracking the Remaining House Races. 2022.https://www.
nytimes.com/interactive/2022/11/10/us/elections/results-house-seats-elections-congress.
html?searchResultPosition=1
[NYT4]
The New York Times. Live Election Results: Top Races to Watch. 2022.https://www.nytimes.
com/interactive/2022/11/08/us/elections/results-key-races.html?action=click&pgtype=Article&
state=default&module=election-results&context=election_recirc&region=NavBar
[NYT5]
The New York Times. World Cup 2022: How Danmark can advance to the round of
16. 2022.https://www.nytimes.com/interactive/2022/upshot/denmark-world-cup-scores-
standings.html?action=click&algo=identity&block=more_in_recirc&fellback=false&imp_
id=39268242&impression_id=b0d7fac1-705c-11ed-8bf3-b7515042b3b5&index=1&pgtype=
Article&pool=more_in_pools%2Ftheupshot&region=footer&req_id=822058229&surface=
eos-more-in&variant=holdout_more-in
[NYT6]
The New York Times. Coronavirus World Map: Tracking the Global Outbreak. 2021.https:
//www.nytimes.com/interactive/2021/world/covid-cases.html
[NYT7]
The New York Times. What the BA.5 Subvariant Could Mean for the United States.
2022.https://www.nytimes.com/interactive/2022/07/07/us/ba5-covid-omicron-subvariant.
html
[NYT8]
The New York Times. Serena Williams: Charting a Career at the Top. 2022.https://www.
nytimes.com/interactive/2022/08/09/sports/tennis/serena-williams-ranking-tennis-wins.html
[NYT9]
The New York Times. U.S. Has Far Higher Covid Death Rate Than Other Wealthy Coun-
tries. 2022.https://www.nytimes.com/interactive/2022/02/01/science/covid-deaths-united-
states.html
[NYT10]
The New York Times. A Detailed Picture of What’s in the Democrats’ Climate and
Health Bill. 2022.https://www.nytimes.com/interactive/2022/08/13/upshot/whats-in-the-
democrats-climate- health-bill.html
[NYT11]
The New York Times. How Nathan Chen Won Gold in Men’s Figure Skating. 2022.https:
//www.nytimes.com/interactive/2022/02/10/sports/olympics/nathan-chen-jumps.html
[NYT12]
The New York Times. The Dollar Is Extremely Strong, Pushing Down the World. 2022.https:
//www.nytimes.com/interactive/2022/07/16/business/strong-dollar.html
[NYT13]
The New York Times. Where $5 Trillion in Pandemic Stimulus Money Went. 2022.https://
www.nytimes.com/interactive/2022/03/11/us/how-covid-stimulus-money-was-spent.html
[NYT14]
The New York Times. How the New Climate Bill Would Reduce Emissions. 2022.https:
//www.nytimes.com/interactive/2022/08/02/climate/manchin-deal-emissions-cuts.html
[NYT15]
The New York Times. Vast New Study Shows a Key to Reducing Poverty: More Friendships
Between Rich and Poor. 2022.https://www.nytimes.com/interactive/2022/08/01/upshot/
rich-poor-friendships.html
[NYT16]
The New York Times. The Exceptionally American Problem of Rising Roadway Deaths.
2022.httpswww.nytimes.com20221127upshotroad-deaths-pedestrians-cyclists.html
[NYT17]
The New York Times. See Everything the White House Wanted, and Everything It Got.
2022.https://www.nytimes.com/interactive/2022/10/20/upshot/biden-budget-before-after-
animation.html
[NYT18]
The New York Times. ‘The Cash Monster Was Insatiable’ How Insurers Exploited Medicare
for Billions. 2022.https://www.nytimes.com/2022/10/08/upshot/medicare-advantage-fraud-
allegations.html